{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,29]],"date-time":"2026-05-29T07:01:40Z","timestamp":1780038100814,"version":"3.53.1"},"reference-count":40,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,5,29]],"date-time":"2026-05-29T00:00:00Z","timestamp":1780012800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:sec>\n                    <jats:title>Introduction<\/jats:title>\n                    <jats:p>Visual grounding aims to localize target objects in images based on given textual descriptions, with broad applications in fields such as autonomous driving and human-robot interaction. However, existing visual grounding models still face three major challenges: (1) Most prior works employ separate encoders to process images and text independently, which enlarges the semantic gap between visual and textual features; (2) The use of large-language models leads to excessive parameters, making deployment on lightweight devices difficult; (3) Single-level cross-modal attention mechanisms are insufficient for fully capturing interactive information across modalities.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>To address these issues, this paper proposes a Task-aware Liquid Cross-modal Network (TLCN), which consists of four key modules: a Feature Extraction Module (FEM), a Liquid Fusion Module (LFM), a Task-aware Cross-modal Refinement Module (TCRM), and a Multilevel Grounding Module (MGM). Specifically, the FEM utilizes textual features to guide the extraction of visual features, thereby reducing the feature gap. The LFM employs Liquid Neural Networks (LNNs) to capture temporal dependencies and significantly reduce model parameters. Furthermore, the TCRM deepens textual representation via a second-level attention mechanism, while designed Conv-Trans Blocks (CTBs) are applied to image data to extract deeper visual features. Additionally, a similarity loss function based on KL divergence is introduced to optimize the cross-modal alignment.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>The proposed model is extensively evaluated on three widely-used public benchmarks: RefCOCO, RefCOCO+, and RefCOCOg. Moreover, a specialized text localization task is designed for further evaluation. Experimental results demonstrate that the TLCN achieves superior performance across all evaluated datasets and tasks.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Discussion<\/jats:title>\n                    <jats:p>The superior performance of TLCN validates the effectiveness of its structural designs: text-guided visual extraction successfully bridges the semantic gap, the introduction of LNNs effectively reduces parameter counts for lightweight deployment, and the second-level attention with CTBs sufficiently captures deep cross-modal interactions. These findings suggest that TLCN provides a promising, efficient, and lightweight solution for visual grounding and related localization tasks.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/frai.2026.1798247","type":"journal-article","created":{"date-parts":[[2026,5,29]],"date-time":"2026-05-29T06:02:42Z","timestamp":1780034562000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Task-aware cross-modal refinement and liquid fusion for text-visual grounding"],"prefix":"10.3389","volume":"9","author":[{"given":"Zhirong","family":"Li","sequence":"first","affiliation":[{"name":"Criminal Investigation Police University of China","place":["Shenyang, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Changliang","family":"Wang","sequence":"additional","affiliation":[{"name":"Shanghai Key Laboratory of Crime Scene Evidence","place":["Shanghai, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yongheng","family":"Pang","sequence":"additional","affiliation":[{"name":"Criminal Investigation Police University of China","place":["Shenyang, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Junchang","family":"Xin","sequence":"additional","affiliation":[{"name":"Northeastern University","place":["Shenyang, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jiaming","family":"Zhang","sequence":"additional","affiliation":[{"name":"Zhejiang University","place":["Hangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yunhe","family":"Sun","sequence":"additional","affiliation":[{"name":"Shenyang Aerospace University","place":["Shenyang, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,5,29]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"107708","DOI":"10.1016\/j.engappai.2023.107708","article-title":"Multimodal emotion recognition via convolutional neural networks: comparison of different strategies on two multimodal datasets","volume":"130","author":"Bilotti","year":"2024","journal-title":"Eng. Appl. Artif. Intellig"},{"key":"B2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52734.2025.00372","article-title":"\u201cTask-aware cross-modal feature refinement transformer with large language models for visual grounding,\u201d","author":"Chen","year":"2025","journal-title":"2025 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)"},{"key":"B3","doi-asserted-by":"publisher","first-page":"7992","DOI":"10.1109\/TNNLS.2022.3148542","article-title":"Recurrent neural networks are universal approximators with stochastic inputs","volume":"34","author":"Chen","year":"2023","journal-title":"IEEE Transactions on Neural Networks and Learning Systems"},{"key":"B4","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51701.2025.00663","article-title":"\u201cPropvg: End-to-end proposal-driven visual grounding with multi-granularity discrimination,\u201d","author":"Dai","year":"2025","journal-title":"Proceedings of the IEEE\/CVF International Conference on Computer Vision"},{"key":"B5","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00179","article-title":"\u201cTransvg: end-to-end visual grounding with transformers,\u201d","author":"Deng","year":"2021","journal-title":"2021 IEEE\/CVF International Conference on Computer Vision (ICCV)"},{"key":"B6","doi-asserted-by":"publisher","first-page":"13636","DOI":"10.1109\/TPAMI.2023.3296823","article-title":"Transvg++: end-to-end visual grounding with language conditioned vision transformer","volume":"45","author":"Deng","year":"2023","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell"},{"key":"B7","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.01601","article-title":"\u201cVision-language transformer and query generation for referring segmentation,\u201d","author":"Ding","year":"2021","journal-title":"2021 IEEE\/CVF International Conference on Computer Vision (ICCV)"},{"key":"B8","doi-asserted-by":"publisher","first-page":"992","DOI":"10.1038\/s42256-022-00556-7","article-title":"Closed-form continuous-time neural networks","volume":"4","author":"Hasani","year":"2022","journal-title":"Nat. Mach. Intell"},{"key":"B9","article-title":"Liquid time-constant networks","author":"Hasani","year":"2020","journal-title":"CoRR, abs\/2006.04439"},{"key":"B10","article-title":"Modeling relationships in referential expressions with compositional modular networks","author":"Hu","year":"2016","journal-title":"CoRR, abs\/1611.09978"},{"key":"B11","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.01661","article-title":"\u201cLook before you leap: learning landmark features for one-stage visual grounding,\u201d","author":"Huang","year":"2021","journal-title":"2021 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)"},{"key":"B12","doi-asserted-by":"publisher","first-page":"102725","DOI":"10.1016\/j.inffus.2024.102725","article-title":"Atcaf: attention-based causality-aware fusion network for multimodal sentiment analysis","volume":"114","author":"Huang","year":"2025","journal-title":"Inf. Fusion"},{"key":"B13","doi-asserted-by":"publisher","first-page":"111346","DOI":"10.1016\/j.knosys.2023.111346","article-title":"TMBL: transformer-based multimodal binding learning model for multimodal sentiment analysis","volume":"285","author":"Huang","year":"2024","journal-title":"Knowl.-Based Syst"},{"key":"B14","doi-asserted-by":"publisher","first-page":"642","DOI":"10.1038\/s42256-020-00237-3","article-title":"Neural circuit policies enabling auditable autonomy","volume":"2","author":"Lechner","year":"2020","journal-title":"Nat. Mach. Intell"},{"key":"B15","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3587251","article-title":"Transformer-based visual grounding with cross-modality interaction","volume":"19","author":"Li","year":"2023","journal-title":"ACM Trans. Multimedia Comput. Commun. Appl"},{"key":"B16","doi-asserted-by":"publisher","first-page":"121515","DOI":"10.1016\/j.ins.2024.121515","article-title":"A deep spatiotemporal interaction network for multimodal sentimental analysis and emotion recognition","volume":"690","author":"Li","year":"2025","journal-title":"Inf. Sci"},{"key":"B17","doi-asserted-by":"publisher","first-page":"128940","DOI":"10.1016\/j.neucom.2024.128940","article-title":"Multimodal sentiment analysis based on disentangled representation learning and cross-modal-context association mining","volume":"617","author":"Li","year":"2025","journal-title":"Neurocomputing"},{"key":"B18","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00477","article-title":"\u201cLearning to assemble neural module tree networks for visual grounding,\u201d","author":"Liu","year":"2019","journal-title":"The IEEE International Conference on Computer Vision (ICCV)"},{"key":"B19","doi-asserted-by":"publisher","first-page":"110847","DOI":"10.1016\/j.patcog.2024.110847","article-title":"Sentiment analysis based on text information enhancement and multimodal feature fusion","volume":"156","author":"Liu","year":"2024","journal-title":"Pattern Recognit"},{"key":"B20","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.01005","article-title":"\u201cMulti-task collaborative network for joint referring expression comprehension and segmentation,\u201d","author":"Luo","year":"2020","journal-title":"2020 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)"},{"key":"B21","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52734.2025.00843","article-title":"\u201cGes3vig: incorporating pointing gestures into language-based 3d visual grounding for embodied reference understanding,\u201d","author":"Mane","year":"2025","journal-title":"2025 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)"},{"key":"B22","article-title":"Generation and comprehension of unambiguous object descriptions","author":"Mao","year":"2015","journal-title":"CoRR, abs\/1511.02283"},{"key":"B23","article-title":"Exploring liquid neural networks on loihi-2","author":"Pawlak","year":"2024","journal-title":"ArXiv, abs\/2407.20590"},{"key":"B24","doi-asserted-by":"publisher","first-page":"1054","DOI":"10.1109\/TKDE.2025.3639419","article-title":"Exploiting simd-ified bit-parallelism for high-performance complex event matching","volume":"38","author":"Qiu","year":"2025","journal-title":"IEEE Trans. Knowl. Data Eng"},{"key":"B25","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01045","article-title":"\u201cLanguage adaptive weight generation for multi-task visual grounding,\u201d","author":"Su","year":"2023","journal-title":"2023 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)"},{"key":"B26","doi-asserted-by":"publisher","first-page":"129842","DOI":"10.1016\/j.neucom.2025.129842","article-title":"Conv-enhanced transformer and robust optimization network for robust multimodal sentiment analysis","volume":"634","author":"Sun","year":"2025","journal-title":"Neurocomputing"},{"key":"B27","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00206","article-title":"\u201cNeighbourhood watch: referring expression comprehension via language-guided graph attention networks,\u201d","author":"Wang","year":"2019","journal-title":"2019 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)"},{"key":"B28","doi-asserted-by":"publisher","first-page":"129163","DOI":"10.1016\/j.neucom.2024.129163","article-title":"Transformer-based correlation mining network with self-supervised label generation for multimodal sentiment analysis","volume":"618","author":"Wang","year":"2025","journal-title":"Neurocomputing"},{"key":"B29","doi-asserted-by":"publisher","first-page":"109174","DOI":"10.1016\/j.engappai.2024.109174","article-title":"A dual-channel multimodal sentiment analysis framework based on three-way decision","volume":"137","author":"Wang","year":"2024","journal-title":"Eng. Appl. Artif. Intell"},{"key":"B30","doi-asserted-by":"publisher","first-page":"387","DOI":"10.1007\/978-3-030-58568-6_23","article-title":"\u201cImproving one-stage visual grounding by recursive sub-query construction,\u201d","author":"Yang","year":"2020","journal-title":"Computer Vision"},{"key":"B31","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01506","article-title":"\u201cShifting more attention to visual backbone: Query-modulated refinement networks for end-to-end visual grounding,\u201d","author":"Ye","year":"2022","journal-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)"},{"key":"B32","article-title":"Modeling context in referring expressions","author":"Yu","year":"2016","journal-title":"CoRR, abs\/1608.00272"},{"key":"B33","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01949","article-title":"\u201cVisual programming for zero-shot open-vocabulary 3d visual grounding,\u201d","author":"Yuan","year":"2024","journal-title":"2024 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)"},{"key":"B34","doi-asserted-by":"publisher","first-page":"115149","DOI":"10.1016\/j.knosys.2025.115149","article-title":"Ta-oem : multimodal sentiment analysis using two-step hierarchical attention and self-supervised label generation","volume":"334","author":"Zhang","year":"2026","journal-title":"Knowl.-Based Syst"},{"key":"B35","doi-asserted-by":"publisher","first-page":"3291","DOI":"10.1109\/TAFFC.2025.3591567","article-title":"Attsf-net: attention-based similarity fusion network for audio-visual emotion recognition","volume":"16","author":"Zhang","year":"2025","journal-title":"IEEE Trans. Affect. Comput"},{"key":"B36","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01241","article-title":"\u201cTowards clip-driven language-free 3d visual grounding via 2D-3D relational enhancement and consistency,\u201d","author":"Zhang","year":"2024","journal-title":"2024 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)"},{"key":"B37","doi-asserted-by":"publisher","first-page":"1523","DOI":"10.1109\/TNNLS.2022.3183827","article-title":"Word2pix: word to pixel cross-attention transformer in visual grounding","volume":"35","author":"Zhao","year":"2024","journal-title":"IEEE Trans. Neural Netw. Learn. Syst"},{"key":"B38","doi-asserted-by":"publisher","first-page":"103296","DOI":"10.1016\/j.inffus.2025.103296","article-title":"Context-driven and sparse decoding for remote sensing visual grounding","volume":"123","author":"Zhao","year":"2025","journal-title":"Information Fusion"},{"key":"B39","doi-asserted-by":"publisher","first-page":"598","DOI":"10.1007\/978-3-031-19833-5_35","article-title":"\u201cSeqtr: a simple yet universal network for visual grounding,\u201d","author":"Zhu","year":"2022","journal-title":"Computer Vision-ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23\u201327, 2022, Proceedings, Part XXXV"},{"key":"B40","doi-asserted-by":"publisher","first-page":"112688","DOI":"10.1016\/j.patcog.2025.112688","article-title":"Dcart: a dual contrastive alignment residual transformer model for visual grounding","volume":"172","author":"Zhu","year":"2026","journal-title":"Pattern Recognit"}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1798247\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,29]],"date-time":"2026-05-29T06:02:43Z","timestamp":1780034563000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1798247\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,29]]},"references-count":40,"alternative-id":["10.3389\/frai.2026.1798247"],"URL":"https:\/\/doi.org\/10.3389\/frai.2026.1798247","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,29]]},"article-number":"1798247"}}